人工智能论文英文版-Recommender systems, stigmergy, and the tyranny of popularity.pdfVIP

Recommendersystems,stigmergy,andthetyrannyof

popularity

ZackaryOkunDunivin1∗

PaulSmaldino2,3

51UniversityofCalifornia,Davis

22

0UniversityofCalifornia,Merced

23SantaFeInstitute

n

u

JJune9,2025

6

]

Y

C

.Abstract

s

c

[

Scientificrecommendersystems,suchasGoogleScholarandWebofScience,are

1

vessentialtoolsfordiscovery.Searchalgorithmsthatpowerworkthroughstigmergy,a

2collectiveintelligencemechanismthatsurfacesusefulpathsthroughrepeatedengage-

6ment.Whilegenerallyeffective,this“rich-get-richer”dynamicresultsinasmallnum-

1

6berofhigh-profilepapersthatdominatevisibility.Thisessayarguesarguethatthese

0algorithmover-relianceonpopularityfostersintellectualhomogeneityandexacerbates

.

6structuralinequities,stiflinginnovativeanddiverseperspectivescriticalforscientific

0progress.Weproposeanoverhaulofsearchplatformstoincorporateuser-specificcali-

5

2bration,allowingresearcherstomanuallyadjusttheweightsoffactorslikepopularity,

:recency,andrelevance.Wealsoadviseplatformdevelopersonhowwordembeddings

v

iandLLMscouldbeimplementedinwaysthatincreaseuserautonomy.Whileour

Xsuggestionsareparticularlypertinenttoaligningrecommendersystemswithscientific

r

avalues,th